Assessing the Payoff for Immigrants

(1)

Assessing the Payoff for Immigrants

and Minorities'

John R. Logan Richard D. Alba

University at Albany

Brian J. Stults University of Florida

Self-employment and work in sectors with high concentrations of owners and workers of the same ethnicity have been identified as potential routes of economic success for immigrants. This study uses 1990 census data to assess the effects of self-employment, ethnic employment, and their interaction on the odds of being at work, on number of hours worked, and on earnings of individual members of several representative groups. These groups include Cubans in Miami; African Americans, Puerto Ricans, Kore- ans, Chinese and Dominicans in New York; and African Americans, Kore- ans, Chinese, Mexicans and Salvadorans in Los Angeles. Work in ethnic

sectors of the economy has no consistent effects, although work in their niche in the public sector offers greater rewards than any other type of employment for African Americans and Puerto Ricans. Findings are mixed for self-employment, and its estimated effect on earnings depends on model specification. We conclude that the self-employed work longer hours but in many cases at lower hourly rates. The effects of self-employment are the same in ethnic sectors as in the mainstream economy.

In the 1960s, a simple dual economy model was a useful device for referring to the concentration of minority workers in certain low-wage industries (Edwards, Reich and Gordon, 1975). The revival of large-scale immigration has highlighted another feature of the twentieth century metropolis: the immigrant proclivity toward small-scale business enterprise as an alternative 'This research was supported by the National Science Foundation (grant number SBR95- 07920) and by the Lewis Mumford Center for Comparative Urban and Regional Research.

The Center for Social and Demographic Analysis, University at Albany, provided technical and administrative support through grants from NICHD (P30 HD32041) and NSF (SBR- 9512290). We thank Michael Dill for his assistance. Comments should be directed to John R.

Logan, Department of Sociology, University at Albany, Albany, NY 12222

<[email protected]>.

(2)

source of livelihood. Select immigrant minorities have secured strong posi- tions as business owners or self-employed workers in certain economic sectors. This is true of the Cubans in Miami (Wilson and Portes, 1980). In New York and Los Angeles, which are the principal centers of immigration in the country, such entrepreneurialism is the basis for some of the largest and most diversified ethnic economies in the country (Logan, Alba, Dill and Zhou, 2000). Two of these have been the subject of well known case studies: the Koreans in Los Angeles (Light and Bonacich, 1988) and Chinese in New York (Zhou, 1992).

We ask what are the impacts of such ethnic economies for the people who are employed in them. More specifically and relevant to current policy debates, do the ethnic enclaves of certain new immigrant groups provide them opportunities to work more steadily or at higher wages than do those African Amer- ican and Hispanic minorities who remain largely confined to employment niches? Does being a business owner or being self-employed, particularly within the ethnic economy, improve people's labor market outcomes?

THE FORMS OF ETHNIC ECONOMIES

Clearly ethnic economies come in many shapes and sizes (for a critical review of related concepts, see Light et al., 1994). The general phenomenon of ethnic clustering in certain parts of the metropolitan labor force is well known, but it can have a variety of sources and consequently develop along various paths. Early immigrants from a particular place of origin may simply discov- er job opportunities in certain jobs, then generate chain migration into those jobs through ethnic and family social networks. Or they may be recruited directly for certain jobs, sometimes by labor contractors seeking workers in a specific country of origin. Opportunities for small business are sometimes created by the growth of an immigrant community to serve that community in retail and service sectors. Less often, perhaps, entrepreneurs can take advantage of a low-wage, co-ethnic labor force to compete in labor-intensive

manufacturing or other activities.

We emphasize the heterogeneity of ethnic economies because their benefits may vary according to the form that they take, and benefits may also be different for workers than for entrepreneurs. As Waldinger (1996b:449-450) concludes in the case of Los Angeles, "Niching is pervasive, but not every niche proves rewarding. Some do, notably those concentrations that provide opportunities for self-employment .... [F]or African Americans, government is an advantageous niche .... By contrast, Mexicans and Central Americans

(3)

seem to have been herded into niches that constitute mobility traps." Fol- lowing this reasoning, what matters is not whether a group concentrates in an ethnic economy, but what kind of ethnic economy the group is able to estab- lish. Unfortunately, little previous investigation of labor market outcomes has distinguished between different kinds of ethnic economic incorporation.

From the preceding quotation, one might infer that Waldinger expects niches in government or with high levels of self-employment to be advantageous, but other niches to be "mobility traps." But his empirical work (Waldinger, 1996a, b) studies the effects of all niches, without distinguishing whether group members tend to work for co-ethnic entrepreneurs or for others or whether they are self-employed. Portes and Zhou (1996), as another example, focus on the effects of self-employment, without considering whether entrepreneurship is clustered into ethnic concentrations or in sectors where owners are likely to be able to employ co-ethnic workers. Our purpose is to link labor market outcomes simultaneously to the form of ethnic concentration (its mix of ownership and labor) and the person's position within it (as an

owner or worker).

We focus on three patterns that have been widely recognized and that

we call employment niches, ethnic enclaves, and entrepreneurial niches

(Logan and Alba, 1999). We identify these patterns based partly on spatial concentration (they apply to specific metropolitan areas), but more specifically on group members' clustering in certain economic sectors as owners, as

workers, or as both owners and workers.2

There are other perspectives from which ethnic economies could be use- fully analyzed. Sectoral clustering is, however, the only basis that we are aware of by which ethnic patterns can be operationalized with available census data.

In the following section, we define the three patterns (summarized in Figure I); note how they have been discussed by other researchers, and review what is known about the effects of each type on individuals' labor market out-

comes.

Employment niches are economic sectors (defined in our research as industries) where group members are disproportionately represented in the labor force, either in public sector jobs or in private businesses that are typi- cally owned and managed by whites or members of another ethnic group.

2There is a necessary ambiguity in the term "owner" - our data source does not allow us to determine whether a self-employed person actually employs other workers. Therefore, we prefer to refer to this phenomenon as the effect of self-employment, which is measured, rather than ownership, which is the variable that many social scientists have in mind.

(4)

Figure I. Four Categories of Economic Sectors

Concentration of workers Pattern of ownership 1. Employment niche Ethnic workers Public or nonethnic owners 2. Enclave economy Ethnic workers Ethnic owners

3. Entrepreneurial niche Nonethnic workers Ethnic owners 4. Nonethnic sectors Nonethnic workers Nonethnic owners

Gold and Light (1998) refer to this as the "ethnic-accessed economy." Lieber- son (1980) noted that members of ethnic groups have often congregated in similar jobs through control of labor unions, information about openings, or other privileged participation in labor recruitment (see also Model, 1993;

Waldinger 1996a). Public jobs have been a particularly important sort of niche throughout this century because of their overall growth and their relative security of employment. Niches in sectors controlled by powerful craft unions have also offered advantages to group members. On the other hand, disproportionate representation of a group in the labor force of a particular industry can also reflect the group's lack of resources, a ghettoization into undesirable jobs. Jiobu (1988:356) calls this situation "ethnic saturation." He posits that it "increases the likelihood that minority individuals will find employment." But where the group lacks control over hiring, firing and business strategy - either due to absence of ethnic ownership or weakness of orga- nized labor - it may result in "low pay, limited upward mobility, little job

security, and episodic employment."

Waldinger's (1996a: 100) analysis of the earnings payoff for working in a group's employment niche in New York in 1990 shows wide variations among groups (as noted above, however, Waldinger defined employment niches without distinguishing between owner and worker concentrations).

Italians, whose niche was mainly in professional and technical sectors by this time, and African Americans, concentrated in public jobs, did better in their niches than outside of them. New immigrant groups, including Chinese, Dominicans and West Indians, earned considerably less in their niche jobs, even after controlling for background characteristics. His analysis of groups in Los Angeles (Waldinger, 1996b) provides similarly variable results.

Wilson (1997) offers another analysis of the effects of employment niches (defined similarly to Waldinger's study, but taking into account both industry and occupation) on labor market outcomes, pooling data from the 1980 and 1990 PUMS samples for the largest 23 metropolitan areas. Again, working in a group's employment niche (categorized very broadly as niches of whites, African Americans, Hispanics, and Asians) has variable effects on

(5)

wages: positive for whites and Hispanics but negative for African Americans and Asians.

Another ethnic pattern is where group members are concentrated as both owners and workers in certain activities, which we call ethnic enclaves (Logan, Alba and McNulty, 1994). With minor differences, this is what Jiobu (1988) calls "ethnic hegemony," where a sheltered ethnic labor market is combined with ethnic economic control; Light (1996) calls this simply the "ethnic economy." The "middleman minority" business owner, serving another group's consumer market but preferring co-ethnic workers (Bonacich,

1973; Zenner, 1991), is a limited form of ethnic enclave. A broader notion has been developed since the term was introduced by Wilson and Portes (1980): an enclave is a complex of economic sectors, perhaps interrelated

among themselves and with a spatially concentrated core area, controlled

through ownership by members of an ethnic group who rely especially on a co-ethnic labor force. We lack information on relations among sectors, flows between ethnic and non-ethnic businesses, and fine-grained geography. Most important, we cannot evaluate the extent to which members of an ethnic

group are employed by other group members. Our operationalization relies simply on group concentrations in particular sectors as owners and workers, presuming that ownership implies control.

Some authors believe that enclaves provide benefits to both owners and workers. Bailey and Waldinger (1991) suggest that from the employer's perspective a co-ethnic labor force provides assurance that investments in training will be repaid by loyalty and (in one form or another) reduced labor costs.

Wilson and Portes (1980) emphasize the advantages to workers: within a sheltered ethnic economy, workers may find employment despite their deficits (such as poor English, lack of formal education, or unfamiliarity with the labor market), while those with better qualifications are more likely to find jobs commensurate with their skills. They may therefore be able to earn

more than comparable workers outside the enclave.

Despite speculation about such benefits, there is little supportive evidence to date. Wilson (1997) finds that working in a sector with a higher share of co-ethnic owners and managers - a situation similar to working in an ethnic enclave sector as we define it - has no effect on the odds of joblessness.

The wage payoff from an enclave job is also uncertain. Logan, Alba and McNulty (1994) report that most enclaves in 1980 consisted of a thin clus- ter of economic sectors with low levels of investment and low average wages.

Zhou's generally positive portrait of New York's Chinatown acknowledges

(6)

that "relative to the mainstream economy, the enclave economy, as a whole, represents the basic characteristics of the broader competitive sector," including limited earnings (Zhou, 1992:118). Consistent with these observations, Wilson (1997) reports that working in a sector with more co-ethnic owners and managers is associated with lower hourly wages for whites, Hispanics and Asians. Finally, Model (1997) finds that the effects vary by sector and over time. She notes that some enclave sectors in New York from 1940-1970 provided higher than average earnings to employees. This is the case of government in 1940 and 1950, when it was an Irish sector, and apparel manufacturing in 1940 and 1950, when it was Russian. But these sectors offered no income advantage by 1970, when they had passed over to other groups (African Americans and Italians, respectively). And some other sectors (e.g.,

retail trade) were never advantageous.

A third situation is sectors where the group predominates as owners and self-employed, but without relying particularly on co-ethnic workers (that is, some group members may be employees in these sectors, but the group is overrepresented only as owners). We refer to these as entrepreneurial niches.

Another is a former enclave sector where group members were initially present as both owners and workers, but where the paid workforce has under- gone an ethnic transition. Light (1996) uses the term "immigrant economy"

to describe the special case when immigrant entrepreneurs from one group recruit labor from other, less entrepreneurial immigrant groups. Such cases are acknowledged in the literature (for example, Iranians in Los Angeles studied by Light et al., 1994), but little has been said about how they might affect labor market outcomes.

More interest has been shown in a related question: whether the entrepreneurial activity on which an enclave economy or entrepreneurial niche is based is well rewarded. Most immigrant or minority small businesses operate with relatively low capitalization, relying in part for their profitability on the long hours that self-employed people are willing to commit to work (i.e., self- exploitation). Self-employment may often represent a second-best option for immigrants whose chances for employment in the mainstream labor market are weak (Light and Rosenstein, 1995). For these reasons, one might expect

self-employment to offer poor - though perhaps reliable - returns. But if we took into account the obstacles to a better job, such as recent immigration and poor English language ability, we might find that this is a better option than working for others. For these reasons, there is much debate on whether self-employment improves annual earnings (for a more negative view, see

(7)

Borjas, 1990; Bates, 1997; more positive results are reported by Waldinger, 1996b:451). Portes and Zhou (1996) offer evidence that the conclusion depends on how much weight is given to outlying cases, to very high earners who are more likely to be self-employed. If the log of hourly wages is pre- dicted, self-employment has a negative effect. For non-logged hourly wages, the effect is positive. In other words, somewhat like a lottery, self-employment has a poor average payoff but a high potential one.

We add a new dimension to this question: whether the effects of self- employment are the same in ethnic and non-ethnic sectors of the metropolitan economy. Scholars who disagree on whether the ethnic economy benefits workers (compare Portes and Jensen, 1987 with Sanders and Nee, 1987) agree that location in the enclave economy is advantageous for owners. Rec- ognizing that there are ethnic business owners outside of the ethnic economy, we test this proposition directly.

Our review of concepts and past research provides a basis for several pre- liminary research hypotheses. We state these from the perspective of the positive effects of ethnic concentrations and self-employment as posited in the work of Portes, supplemented by Waldinger's positive evaluation of the public employment option for minorities.

H1. The best outcomes for ethnic minorities, whether self-employed or employees of others, are found in enclave sectors and entrepreneurial niches.

For self-employed this is due to their ability to draw on ethnic business networks and ethnically-based labor recruitment and retention. Employees benefit from privileged access to jobs, training and promotion.

H2. That public employment provides better outcomes than private

sector employment niches or the mainstream economy, though not necessar-

ily better than work in an enclave sector or entrepreneurial niche. The sources of benefit are civil service protection (and in some locales also union mem- bership) and the public sector's application of bureaucratic norms in promotion and pay. Of the groups in this study, only African Americans and Puer- to Ricans have such niches, presumably because they benefit from affirmative action policies. For these groups, one might think of public employment as a form of ethnic enclave.

H3. Self-employment offers members of minority ethnic groups more steady work and higher earnings than working for others. The advantages are the potential for longer working hours, ability to draw on ethnic and family networks, and flexibility in making use of one's full array of abilities.

H4. The benefits of self-employment are higher in the ethnic economy.

The special sources of benefit in enclave sectors are the greater relevance of

(8)

ethnicity to business connections and to labor recruitment and retention. The self-employed in enclaves are expected to benefit from both of these. Those in entrepreneurial niches and those (few) in employment niches may benefit from only one or the other. By contrast, self-employed persons outside the ethnic economy may derive no benefit, but only costs, from their ethnicity.

We evaluate these hypotheses with respect to three different outcomes.

If there were consistent evidence supporting any hypothesis for all three - being at work, hours worked, or wages - the hypothesis would be fully con- firmed. Because we study several groups in two different regions, it would be surprising if any hypothesis were supported for all groups and all outcomes.

Rather, we will look for a pattern in which similar results are found for more

than one predictor, more than one outcome, or more than one group.

RESEARCH DESIGN

The first question of research design - whom to study - carries many impli- cations. For simplicity one would prefer a very limited number of groups, but for generalizability it is necessary to have both ethnic and geographic vari- ation. Because of its importance as a prototype for the concept of economic enclave, we begin with an examination of Cubans in the Miami-Hialeah metropolitan area (to maximize sample size, we study the CMSA, including Ft.

Lauderdale). Then we turn to several of the largest immigrant and minority groups in the New York and Los Angeles-Long Beach metropolitan areas (PMSAs). Koreans and Chinese are included in both metropolitan regions as representatives of the strongest nonwhite ethnic economies. To represent large immigrant groups believed to have weaker ethnic economies, we include Dominicans in New York and Mexicans and Salvadorans in Los Angeles.

Finally, to provide comparative information on mainly nonimmigrant minority groups, we study African Americans in both regions and Puerto Ricans in New York.

We rely on census categories of race to identify African Americans (selecting only those who are non-Hispanic), Koreans and Chinese; we rely on categories of Hispanic origin to identify Puerto Ricans, Mexicans,

Dominicans and Salvadorans. Samples include both immigrants and persons born in the United States, and nativity is included as a control variable in all the multivariate models (though not reported in Tables 5-8). The relevance of this variable differs among groups. Among Chinese, Koreans, Dominicans and Salvadorans, no more than 3 percent of the sample is U.S. born; most are immigrants from before 1985. Among Mexicans and Puerto Ricans, 30-40

(9)

percent of men and a slightly higher share of women were born in the Unit- ed States (and not in Puerto Rico). There is a sizeable share (38% of men and 35% of women) of foreign-born blacks in New York; these are primarily Afro- Caribbeans, who have been shown in some studies to have better labor market success than the U.S. born (Model, 1995; Kalmijn, 1996). The foreign born are less than 5 percent of blacks in our Los Angeles sample.

For each group, we separately estimate models for men and women. Most previous research has considered men only. But as we shall see, women are more than a third of the labor force for all of the groups studied here, and they constitute a majority among African Americans. Therefore it would be a mistake to ignore the situation of women workers. Zhou (1992) found that human capital returns for Chinese immigrant women in New York were much lower than for men; it is plausible that comparable gender differences might be found in returns to employment in ethnic sectors or to self-employment.

Because we will estimate so many equations and make so many com- parisons across groups (and by gender), it may be helpful to state at the begin- ning how we will interpret the results. The key cases for much of the ethnic economy literature are the Cubans, Koreans and Chinese. It is for them, if for anyone, that location in an ethnic sector, and/or self-employment, should have positive returns. We will look for a pattern of statistically significant effects for any one of these groups, starting with Cubans, for both men and women, in any region. We will then turn to the other groups (African Amer- icans, Puerto Ricans, Mexicans, Dominicans and Salvadorans) to see whether their experiences are consistent with, or contradictory to, results for the first

set.

DEPENDENT VARIABLES

We analyze three outcomes of interest: whether group members are currently working; the number of hours that they work; and their annual earnings.

Information on the distribution of these variables for all groups studied here is provided in Table 1.

Current work status is important because the minimal positive claim for ethnic economies is that they provide jobs to group members who might otherwise be unemployed. To study whether people currently work, we begin with a sample of all persons aged 25-64 who have ever worked in the last five years in the civilian labor force. Our choice is guided by characteristics of the

1990 census data on which we rely. The census probes for people's "class of worker" (i.e., self-employed, etc.) and industry, even if they are currently not

(10)

TABLE 1 SAMPLE SIZES AND CHARACTERISTICS

Miami New York Los Angeles

African Puerto African

Cuban Chinese Korean American Rican Dominican Chinese Korean American Mexican Salvadoran

Males Age 25-64a 164,810 74,120 23,584 458,587 202,175 79,094 68,547 39,691 234,402 592,606 64,387 Worked in last 5 years 156,799 69,823 22,249 408,175 170,891 69,950 61,980 36,947 205,398 562,818 62,124 Worked last week 137,202 59,746 19,722 308,803 130,473 54,108 53,897 33,025 154,312 479,145 52,831 % worked last weekb 87.5% 85.6% 88.6% 75.7% 76.3% 77.4% 87.0% 89.4% 75.1% 85.1% 85.0% Earnings samplec 132,568 57,718 18,066 290,689 123,132 49,720 52,137 30,872 146,351 455,980 49,476 Average hours 2,105 2,104 2,293 1,939 1,975 1,968 2,056 2,168 2,007 1,973 1,895 Average earnings 25,914 23,677 25,139 26,101 24,274 18,909 33,796 33,581 29,804 21,252 14,995 Females Age 25-64a 170,684 75,020 23,142 604,936 252,719 99,718 75,026 44,541 272,469 545,460 73,080 Worked in last 5 years 135,110 62,102 16,916 501,521 148,794 67,685 56,752 32,567 220,735 386,161 58,321 Worked last week 106,893 50,468 13,213 382,889 105,000 42,906 45,542 24,898 166,481 278,489 43,296 % worked last weekb 79.1% 81.3% 78.1% 76.3% 70.6% 63.4% 80.2% 76.5% 75.4% 72.1% 74.2% Earnings samplec 101,506 48,307 11,515 361,634 97,877 39,006 42,972 21,946 157,410 257,323 39,142 Average hours 1,882 1,943 2,028 1,826 1,753 1,747 1,860 1,992 1,871 1,773 1,728 Average earnings 17,467 19,133 21,282 22,624 19,650 14,247 23,487 22,153 23,521 15,964 10,310 apopulation age 25-64 less military employed. bPercent that worked last week of those who worked in the last 5 years. cAge 25-64, worked at least 160 hours last year, and had a positive income last year.

(11)

in the labor force, if they have worked in the last five years. Because these characteristics are essential to our analysis, people with no work history, because they never desired to work or could never find work, are excluded.

Among those who have ever worked in the last five years, we predict whether the person was employed and at work in the previous week. Those not at work could have been employed but not working in that particular week, or they could have been unemployed or no longer in the labor force. Because the definitions of these latter categories may not always be clear to respondents, and because in many cases information was reported about people by another household member, we consider the dichotomy of "at work" vs. "not at work" to be the most reliable treatment of this variable. This operationalization catches the unemployed, the "hidden unemployed" of discouraged former job seekers, and the underemployed (working some weeks but not others) in the same net.

Table 1 lists the total number of men and women in each group within the age range that we study (ages 25-64, not employed by the military) who have worked in the past five years and for whom class of worker and industry are available and the number of the latter group who were at work in the last week (for reference, it also provides the total number of persons in this age group, which allows readers to calculate overall rates of labor force participation). Among men, groups fall roughly into two categories. Only 75-77 percent of African Americans, Puerto Ricans and Dominicans worked in the last week, while 85-89 percent of Chinese, Koreans, Mexicans and Cubans did so. By this measure of "steady work," the latter immigrant groups performed significantly better. Among women, there is a more graduated hierarchy of groups: the lowest rates of work are for Dominicans, Puerto Ricans and Mexicans; African American and Salvadoran women are barely distinguishable from Korean women; and Chinese and Cuban women are slightly more likely to be at work.

Other dependent variables are hours worked and earnings. The benefit of higher earnings is self-evident. Working hours are subject to varying inter- pretations. Our view is that the possibility for a full-time person to work steadier hours around an average of 2,000 hours per year (40 hours per week for 50 weeks) is a benefit; those who cannot find this much work suffer a handicap in earning a living. There are, on the other hand, extremes of over- work, and the combination of long hours at a low hourly wage is not neces- sarily desirable.

Hours worked is an annual total. Earnings are the sum of wage/salary

(12)

and self-employment income in 1989. Like Portes and Zhou's study of earnings (1996), our analyses of hours worked and earnings are limited to persons aged 25-64 who worked above a minimal threshold of working hours (160 hours per year) and earned a positive income in the past year. Table 1 shows the size of this population for each group (referred to in the table as the "earnings sample," because it is from this population that we drew the samples for studying working hours and earnings), as well as their mean hours worked and earnings. This sample excludes most retired persons, as well as casual workers and those not in the labor force at all. A small number of people who may be family workers, people working over 160 hours but reporting no income, are also excluded (ranging up to about 4% of Korean women who would otherwise be part of the study).

Korean men in New York and Los Angeles, Chinese men in New York, and Cuban men in Miami have the highest average working hours of any group. Their longer working hours add to the advantage in rates of labor force participation as shown above. However, Chinese and Mexican men in Los Angeles, though they were more likely than most other groups to be at work in the last week, are found to work about the same average annual hours as African American, Puerto Rican, and Dominican men, with Salvadorans not far below. Among women, Koreans also stand out, but African Ameri- can, Chinese and Cuban women work similar hours to one another. It is Puerto Rican, Dominican, Mexican and Salvadoran women who work the least hours.

Given these fairly consistent hierarchies in terms of steady employment, it is surprising that the averages for earnings follow a different order. African American men and women have among the highest annual earnings, compared to other groups. (African American men are the highest earners in New York, and African American women in both regions have the highest earnings compared to women in other groups in the same region; the same finding holds even if Afro-Caribbean blacks are not included). The lowest values, sharply lower than all other groups, are found for Dominicans, Mexicans and Salvadorans. This result is a useful reminder that working more does not nec- essarily mean earning more. Another possibility is that the reported earnings of some groups do not reflect their real earnings, an endemic problem in studies of the self-employed and of groups who participate in underground

economies. Of the dependent variables that we study here, earnings may be the one with the largest share of random or nonrandom error in measure-

ment.

(13)

PREDICTORS

One key independent variable is employment in an ethnic sector of the economy. This is a set of dummy variables based on analyses of each group's sectoral concentrations as owners and/or workers in particular economic sectors.

Overrepresentations are calculated as odds ratios, applying the methodology introduced by Logan, Alba and McNulty (1994). For this purpose, all private-sector workers, male and female, in each metropolitan area (MSA) have been classified by type of worker (owner or self-employed versus employed by somebody else, including unpaid family workers) and by industry sector (a recombination of two-digit industry codes into 47 categories). The odds ratio for "owners" is the ratio of the odds of a group member being an owner or self-employed in a particular sector (versus being an owner or worker in any other sector) to the odds of a non-group member being an owner in this sector (versus being an owner or worker in any other sector). The odds ratio for "workers" is the equivalent ratio for being a worker. These measures have the advantages of being independent of the sizes of groups and of industry sectors, as well as not being affected by the overall distribution of owners across sectors. An odds ratio of 1.00 indicates that a group is neither overrepresented nor underrepresented in a sector. Following current practice, we identify instances in which the value is 1.50 or above (and where the unweighted sam-

ple size for the group in that cell is at least 3) as "concentrations."3

This approach to identifying ethnic sectors is practical, and its results turn out to fit well with fieldworkers' sense of the core sectors of ethnic economies that have been studied more closely. Yet it has inherent limitations.

It counts people as working in an ethnic enclave, based on their industrial sector, who actually may be employed by a member of some other group, working far from the enclave's geographic hub, and for a business that has no con- tacts with enclave firms. Equally, it fails to count as enclave participants some people whose businesses are tightly enmeshed in ethnic networks because their industrial sector is atypical for their group. It would be preferable to identify ethnic sectors based upon intensive field studies in each metropolitan area and to be able to take into account directly the ethnicity of each firm's owners and workforce. Such data are not now publicly available, pre- 3For the exact odds ratios in each sector for the groups studied here, see Logan and Alba et al., 1999. One might wish to use the odds ratios themselves as predictors. It is intuitively appeal- ing to distinguish between very high concentrations and those that barely meet the cutoff. But it is not clear what one would expect from differences in the range below 1.5 or 1.0, or how

to model such effects.

(14)

TABLE 2 DISTRIBUTION ACROSS CATEGORIES OF EMPLOYMENT Miami New York Los Angeles African Puerto African Cuban Chinese Korean American Rican Dominican Chinese Korean American Mexican Salvadoran Males Self-employed 25,800 6,985 4,949 15,314 5,819 4,896 9,628 11,603 11,308 33,445 2,874 Employment sector Ethnic enclave 9,578 18,325 7,324 31,911 -- 6,793 8,498 7,280 18,156 15,752 2,319 Entrepreneurial niche 45,492 4,358 4,057 28,146 -- 4,105 22,412 14,394 14,266 - 242 Employment niche 5,691 747 -- 26,671 10,173 11,525 1,413 -- 26,509 230,897 14,937

Public worker - - - 79,912 27,319 -- - - 36,497 - -

Other 71,807 34,288 6,685 124,049 85,640 27,297 19,814 9,198 50,923 209,331 31,978 Total 132,568 57,718 18,066 290,689 123,132 49,720 52,137 30,872 146,351 455,980 49,476 Females Self-employed 7,671 2,728 2,381 7,654 2,342 1,581 4,292 5,973 5,343 11,374 2,990 Employment sector Ethnic enclave 13,267 16,068 5,298 85,522 - 1,174 9,935 6,523 37,649 1,426 5,158 Entrepreneurial niche 16,762 2,505 2,670 8,523 -- 538 14,242 8,839 8,451 - 1,202 Employment niche 9,327 2,229 - 65,522 5,811 14,232 1,453 - 30,911 74,465 16,013 Public worker -- - - 117,703 27,344 - - - 50,043 - -- Other 62,150 27,505 3,547 84,364 64,722 23,062 17,342 6,584 30,356 181,432 16,769 Total 101,506 48,307 11,515 361,634 97,877 39,006 42,972 21,946 157,410 257,323 39,142 Notes: These cases are age 25-64, worked at least 160 hours last year, and had a positive income last year. - This group not overrepresented in this sector.

(15)

venting us from testing Portes' notion of the ethnic enclave economy more directly.

A different approach is taken to civilian public employment, which does not fit an "owner/worker" categorization. The odds ratio for public employment is the ratio of the odds that a group member is a public employee (compared to an owner or worker in any other sector) to the odds that a non-group member is a public employee (compared to an owner or worker in any other

sector). We consider odds ratios above 1.50 as evidence of an employment niche in the public sector.

Based upon these odds ratios, we define an ethnic enclave to include all industrial sectors where group members are concentrated as both owners and workers. An employment niche is all sectors where they are concentrated only as workers, and an entrepreneurial niche is all sectors where they are concentrated only as owners. A public employment niche exists if the group is concentrated as workers in the public sector. By our measure, public employment is defined as a niche for African Americans in New York and Los Angeles and for Puerto Ricans in New York. We will refer to the "ethnic economy" as the sum of these ethnic sectors and to the remaining sectors as the "mainstream"

(or non-ethnic) economy. Of course, what is "mainstream" for one group may be "ethnic" for another. Table 2 describes the distribution of the labor force across these categories for group members whose hours and earnings are studied here (for reference, the specific industry sectors by category for every group are listed in the Appendix Table; note that in some cases these sectors employ only a small number of group members and therefore have corre- spondingly smaller weight in our analyses).

As has been reported by other researchers, self-employment is extraor- dinarily high for Cuban men in Miami (near 20%), and it is even higher (up to 38%) for Korean men and women in both New York and Los Angeles. It reaches nearly 20 percent for Chinese men in Los Angeles, though it is closer to 10 percent for this group in New York. It approaches 10 percent for Dominican men, but it is substantially lower for African Americans, Puerto Ricans, Mexicans and Salvadorans.

We find that every group has at least a substantial minority of its members working in an ethnic economy as we have defined it, but the specific pattern varies greatly across groups. The main distinction is between those with large enclave and entrepreneurial components and those primarily in public or private employment niches. Compared to all other groups, Chinese and Koreans in New York have the largest shares of employment - a third or more - in their enclave sectors. In Los Angeles these two groups stand out more for

(16)

their entrepreneurial niches. The entrepreneurial niche is also the largest sector for Cuban men (34%), though much smaller for Cuban women. Hence, as we have defined it, the Cuban ethnic economy in Miami is primarily based on entrepreneurship in sectors where Cuban workers are not disproportion-

ately employed.

The shares in enclave or entrepreneurial niches are much lower for other groups; Puerto Ricans have none of either type, while Mexicans have no entrepreneurial niche. One surprising finding is that quite a large share, nearly a quarter, of African American women in New York are in what we have categorized as enclave sectors. On closer inspection, we find that most are in the non-public hospital sector. This case does fall within our definition of an enclave sector, though it is unusual because it has a very large workforce and few self-employed or business owners (indeed, only 1% of African Americans in this sector are classified as self-employed or owners). The hospital sector might be better thought of as a "pseudo-public" sector, because government has at least indirect control of most hospitals in New York. Social services, the other African American enclave sector in New York, has this same character- istic. As we report results for the effect of "enclave sector" employment for African Americans in New York, below, readers may prefer to think of this category as another form of "public" employment.

The less entrepreneurial groups do have large shares in private and public sector employment niches. For African Americans in both regions, these sectors employ between a third and half of the active labor force, more in the public than in the private sector. Puerto Ricans have a similar pattern, though with somewhat smaller percentages. Some other groups - Mexicans, Salvado- rans and Dominicans - have large shares concentrated in private sector employment niches. In contrast, employment niches are of little importance for Koreans, Chinese and Cubans.

In the multivariate models below, four dummy variables represent working in an enclave sector, entrepreneurial niche, employment niche, or in the public sector (included only in the African American and Puerto Rican models). Working elsewhere, referred to as "mainstream sectors," is the refer-

ence category.

Self-employment is measured to include both the self-employed and business owners, in contrast to wageworkers in civilian occupations. We estimate the direct effects of self-employment on being at work, hours worked and earnings. We also test whether self-employment has different effects in different kinds of industrial sectors by introducing a series of interaction

(17)

terms, estimating the effects of self-employment in combination with enclave, entrepreneurial niche and employment niche employment. In some equations, where an interaction term would be based on less than 25 cases as either self-employed or non-self-employed in a type of sector, the coefficients are omitted from the reported results. There is, of course, no interaction term

with public employment.

Other independent variables are introduced as control variables and are coded to match the models estimated by Portes and Zhou (1996). In the following tables, the coefficients for control variables are provided only in the Cuban equations; they were generally the same in equations for other groups.

Marriage is a dummy variable distinguishing those currently married from single, divorced and widowed persons. Living with children is a dummy variable identifying those who live in a household with children under the age of

18. Age is a modification of actual age, used in the literature to impute years of work experience; it is age minus years of schooling. Because many people in our sample may not have worked continuously in their adult years, we refer to this variable simply as age. Occupation is represented by a dummy variable for executive, managerial, administrative or professional occupations,

and another dummy variable is for technician and precision production occupations. Education is represented as a set of dummy variables for some high school, some college and some post-graduate education based on number of years of schooling, with less than eight years as the reference category. Eng- lish language ability is a dummy variable contrasting those who speak English only or well with those who speak English poorly or not at all. Immigration is included as a contrast between the most recent immigrants (those arriving in the previous 5 years), earlier immigrants and those born in the United States (the reference category).

There is disagreement over the most appropriate way to represent earnings as a dependent variable. Our earnings equations use annual earnings and include hours worked as a predictor. This is equivalent to an alternative specification, where the dependent variable is earnings per hour. We prefer this form because it explicitly calls attention to the importance of hours worked as a predictor of total earnings. Portes and Zhou (1996) point out that there may be a question about the causal relation between hours worked, earnings and self-employment (see also Petersen, 1989). Very likely, self-employment increases working hours: "Entrepreneurs may be less constrained than are salaried workers in their choice of work hours and, given a satisfactory return, are willing to put in extra work effort" (Portes and Zhou, 1996:221). Hours

(18)

worked could affect self-employment if people's expectations about the long hours required to run a small business affected their decision of whether to open one. Hours worked could be affected by hourly earnings, though the direction of the effect is unclear: one might be eager to work more hours if the rate was high, but at the same time, one might be forced to work more hours if the rate was low. Similar issues may arise in regard to working in an ethnic sector of the economy. Such work might result in greater hours but lower hourly wages; lower wages might result in working more hours.

Our models assume that self-employment and ethnic employment may affect hours worked and that both of these predictors (and their interaction) and hours may affect total earnings. In future studies using longitudinal data these assumptions could be tested directly.

Another measurement issue is whether to use the logged or non-logged value of earnings as the dependent variable. Portes and Zhou (1996) use both, and we also estimate both forms of the model. There are strong statistical reasons to prefer the logged models: non-logged earnings depart more substantially from normality as indicated by the Jarque-Bera test (Jarque and Bera,

1987). There are also substantive differences between the two model specifi- cations, and these are clearly associated with the importance given to outliers among the self-employed. As an example, among Cubans in our Miami PUMS sample, only 0.6 percent of non-self-employed men reported incomes over $180,000; 2.2 percent of self-employed men had incomes this high (and several were over $300,000). These outliers are theoretically important, because they may be quite influential in their communities. They "can have social and economic effects on their communities that go well beyond pure- ly individual success" (Portes and Zhou, 1996:228). The logged models give less importance to such cases, reflecting the experience of more typical group members. The findings in non-logged models, on the other hand, are very much influenced by outliers. In our presentation of results, if there is a negative effect on logged income but a positive effect on non-logged income, we will describe this combination as an "overall disadvantage" apart from "excep- tional" cases. It would not be accurate, in our view, simply to report that the findings are "mixed" in that event.

The tables present unstandardized regression coefficients. We estimated multivariate models using weighted cases (adjusting the weights so that the final weighted sample was the same size as the initial unweighted sample). In order to compensate for great differences in the size of groups, which influ- ence tests of statistical significance, the final weighted N is limited to 10,000.

We would not wish to accept a null hypothesis for one group and reject it for

(19)

another only because the samples were of vastly different sizes. Still, models for smaller groups such as Koreans and Chinese tend to have higher standard errors for coefficients than do larger groups such as African Americans and Mexicans.

THE CUBAN ENCLAVE IN MIAMI

We begin with a review of the paramount example of an ethnic economy, that of Cubans in Miami. What are the effects of self-employment or location in an ethnic sector on the probability of working, on hours worked and on earnings in this case? The relevant results are presented in Tables 3 (for men) and 4 (for women).

The first column of each table reports a logistic regression model for Cubans who have worked in the last five years, predicting the odds of being at work in the previous week. The probability of working is closely tied to

occupational level; for women, it is also related to higher education and better English language facility. Recent immigrants suffer no disadvantage compared to natives or more established immigrants. The effects of demographic variables depend on gender. Among men, working is positively associated with being married, having children and being younger. Among women, it is negatively associated with marriage and children and unrelated to age. Most relevant to our inquiry, there is no effect of self-employment or of being in an ethnic sector of the economy and no interaction effects for either Cuban men or Cuban women.

The second column of Tables 3 and 4 reports a multiple regression

model predicting hours worked last year (scaled in 100s of hours) for those who worked at least 160 hours and earned at least $500 during the year.

Human capital variables have strong effects in this equation: there are large benefits to higher occupational levels, higher education (except for women) and English language facility. For men, there is a substantial disadvantage to recent immigrants; among women, earlier immigrants work the most hours.

Age has no effect for either gender. Consistent with the previous model, men who are married and have children work longer hours; the opposite effects are found for women.

Self-employment substantially increases working hours (by as much as 90 hours per year) for both men and women. But it is self-employment in general, rather than work in ethnic sectors of the economy, that pays off in more steady employment. There are no main or interaction effects of ethnic sector for men, and there are mixed effects for women (a positive effect of

(20)

TABLE 3 PREDICTING LABOR MARKET OUTCOMES FOR CUBANS IN MIAMI-HwIALAH - MALES

Log-odds of working Hours worked last year Yearly Earnings Yearly Earnings

Miami last week (in 100s) (logged) (non-logged)

Occupation Other Exec-Manage-Admin-Prof. 0.271 (0.131) 0.777 (0.227) 0.248 (0.023) 10064 (776) Technician-Precision Production 0.480 (0.189) 0.842 (0.319) 0.087 (0.032) 853 (1088)

Education Grammar School - Some High School 0.059 (0.115) 0.379 (0.258) 0.086 (0.026) 1221 (880) Some College 0.243 (0.141) 1.078 (0.295) 0.377 (0.030) 8477 (1009) Some post-graduate 0.335 (0.244) 1.600 (0.430) 0.605 (0.044) 22572 (1469)

Speaks English well 0.170 (0.100) 1.080 (0.213) 0.231 (0.022) 5987 (728) Immigration Status U.S.-born Pre-85 immigrant 0.117 (0.190) 0.464 (0.356) 0.053 (0.036) 1642 (1214) Post-85 immigrant -0.203 (0.234) -1.535 (0.494) -0.291 (0.050) -3828 (1686)

Married 0.635 (0.091) 1.551 (0.200) 0.245 (0.020) 5025 (687) HH with children 0.192 (0.091) 0.492 (0.180) 0.050 (0.018) 3017 (614) Age -0.120 (0.041) 0.058 (0.084) 0.027 (0.008) 1152 (286)

Yearly hours (in 100s) 0.047 (0.001) 817 (43)

Self-employed 0.292 (0.161) 0.910 (0.280) 0.014 (0.028) 9088 (956) Enclave -0.159 (0.161) 0.613 (0.347) -0.011 (0.035) -745 (1184) Entrepreneurial niche -0.124 (0.097) -0.330 (0.203) 0.054 (0.021) 1343 (693) Employment niche 0.130 (0.238) -0.279 (0.424) 0.059 (0.043) 211 (1446) Self x Enclave 0.397 (0.606) 0.141 (0.965) -0.293 (0.098) -9719 (3291) Self x Ent. Niche 0.006 (0.230) -0.128 (0.428) -0.137 (0.043) -9612 (1459) Self x Emp. Niche a Public Employee b b b b Constant 1.679 (0.245) 17.531 (0.494) 8.126 (0.055) -13204 (1843)

Goodness of fitc 169.5 .049 .370 .258

7,270 6,449 6,449 6,449 Notes: a The coefficients for these cells are omitted because they are based on less than 25 cases.

b This group not overrepresented in this sector. 'Likelihood ratio for logistic regression model and R-square for OLS models. - Reference category

(21)

TABLE 4 PREDICTING LABOR MARKET OUTCOMES FOR CUBANS IN MLAMI-HIALEAH - FEMALES Log-odds of working Hours worked last year Yearly Earnings Yearly Earnings Miami last week (in 100s) (logged) (non-logged) Occupation Other Exec-Manage-Admin-Prof. 0.438 (0.116) 0.613 (0.222) 0.307 (0.021) 6548 (424) Technician-Precision Production 0.162 (0.160) 0.859 (0.355) 0.121 (0.033) 2133 (675) Education Grammar School - Some High School 0.210 (0.115) -0.238 (0.289) 0.045 (0.027) 409 (550) Some College 0.261 (0.135) 0.157 (0.326) 0.242 (0.031) 3729 (620) Some post-graduate 0.800 (0.257) -0.036 (0.464) 0.500 (0.044) 12053 (882) Speaks English well 0.330 (0.094) 0.513 (0.226) 0.279 (0.021) 4509 (431) Immigration Status U.S.-born Pre-85 immigrant 0.179 (0.159) 0.713 (0.340) 0.030 (0.032) 528 (647) Post-85 immigrant 0.020 (0.214) -0.830 (0.521) -0.260 (0.049) -3063 (992) Married -0.287 (0.082) -0.373 (0.176) -0.004 (0.016) -240 (334) HH with children -0.193 (0.081) -0.879 (0.180) 0.007 (0.017) 106 (344) Age 0.028 (0.041) -0.136 (0.089) -0.048 (0.008) -578 (169) Yearly hours (in 100s) 0.060 (0.001) 679 (27) Self-employed 0.086 (0.164) 0.828 (0.354) -0.090 (0.033) 2573 (675) Enclave -0.123 (0.112) 0.505 (0.277) -0.046 (0.026) -580 (527) Entrepreneurial niche -0.121 (0.106) -0.696 (0.240) 0.029 (0.023) 140 (458) Employment niche 0.178 (0.150) 1.234 (0.298) 0.032 (0.028) -402 (569) Self x Enclave -0.421 (0.431) -1.065 (1.121) -0.066 (0.105) -2273 (2134) Self x Ent. Niche -0.459 (0.389) -0.969 (1.014) 0.041 (0.095) -1164 (1931) Self x Emp. Niche a Public Employee b b b b Constant 1.357 (0.234) 18.513 (0.522) 8.061 (0.055) -1535 (1112) Goodness of fitc 103.2 .023 .473 0.331 6,009 4,982 4,982 4,982 Notes: a The coefficients for these cells are omitted because they are based on less than 25 cases. b This group not overrepresented in this sector. c Likelihood ratio for logistic regression model and R-square for OLS models. - Reference category

(22)

working in the employment niche, but a negative effect of working in the

entrepreneurial niche).

Our analysis of earnings is reported in the third and fourth columns of these tables. Here, as noted above, hours worked in the past year is included as an additional predictor. In the logged models, the effect of each independent variable can be interpreted readily as the percentage increase in earnings associated with a unit change in the predictor. As expected, in both versions

of the earnings model there are strong and significant effects of occupation,

education, immigration and English language facility for both men and women. Older married men and those with children have higher earnings.

Older women earn less, while marital status and living with children have no effect on their earnings. Hours are, of course, a very strong predictor of earnings: every hundred hours worked increases annual earnings by about 5 percent or $700-800.

Our main interest is in the effects of self-employment and ethnic sector.

Here the results depend on the model specification. For logged earnings, the effects are mainly negative. Among women, the outcome is simple: the self- employed earn less, and there are no effects of working in an ethnic sector.

The presence of some significant interaction terms complicates the results for men. First, there is no main effect of self-employment. Second, being in the entrepreneurial niche increases earnings by 5 percent, but this result holds only for the few workers in this category. The large but negative interaction term of entrepreneurial niche with self-employment implies that the self- employed in this sector earn about 10 percent less than those in the general reference category (workers in the mainstream economy). Third, the negative interaction term of enclave sector with self-employment indicates that the self-employed in this sector earn nearly 30 percent less than workers in the enclave or workers/self-employed in the mainstream economy.

In the non-logged model, results for women are reversed: self-employment has a positive effect, while ethnic employment still has no effect. For men, the result again is complicated by interaction effects. For persons in the mainstream economy (the reference category), self-employment offers an advantage in earnings. In both the enclave and the entrepreneurial niches, however, there is no net effect of self-employment (that is, the main effect is counterbalanced by a negative interaction term of equal size).

Taking these results together, we find that there are, if anything, negative effects of being in an ethnic sector of the economy and no disadvantage (indeed, some advantage for women) of the employment niche compared to

(23)

the mainstream economy. Hypothesis 1 is thus not supported. Self-employment is beneficial for Cuban men and women by contributing to total hours worked. In this way it indirectly (but substantially) increases earnings. How- ever the direct effect of self-employment on earnings appears only for excep- tional cases (that is, in the non-logged model), and it does not apply even here to all sectors. For the bulk of Cubans (as shown in the logged model), the effects are negative for women and also negative for men in some sectors.

Hence, Hypothesis 3 is given only mixed support. Finally, the only evidence that the effects of self-employment differ across economic sectors is found in the earnings equations for men. Here the results contradict Hypothesis 4: in both versions of the earnings model, self-employment has the least benefit for men in the enclave or entrepreneurial niche.

IMMIGRANT AND MINORITY GROUPS IN NEW YORK AND LOS ANGELES

We now turn to a larger set of immigrant and minority groups, seeking to generalize beyond the Cuban experience in Miami. Do immigrant groups in New York and Los Angeles - or at least, do some groups such as Koreans and Chinese who are often considered to be especially advantaged by their ethnic economies - get greater benefits from ethnic jobs than do Cubans? Is self- employment beneficial mainly in terms of working hours for these groups, as for Cubans, or is there also a broader payoff in earnings? What is the paral- lel situation for those groups that are generally considered to be disadvantaged in the labor market: African Americans in both regions, Puerto Ricans and Dominicans in New York, and Mexicans and Salvadorans in Los Ange-

les? Does self-employment ever pay off for them? How does work in ethnic sectors of the economy affect them? Is there any advantage from their limited enclave or entrepreneurial sectors? Are they particularly disadvantaged in their private sector employment niches? Does public employment (for African Americans and Puerto Ricans) enhance their outcomes from work?

Results of multivariate models for men and women are presented in Table 5 (for working last week), Table 6 (for hours worked), Table 7 (for logged earnings) and Table 8 (for non-logged earnings). Because of the large number of groups analyzed in this portion of the study, we provide only the coefficients for the key self-employment and ethnic sector variables. Among the control variables, indicators of human capital are almost uniformly positively associated with earnings and, in many cases, also with hours worked.

New immigrants tend to be disadvantaged. The effects of these variables on

(24)

working last week, however, are scattered and in mixed directions. Family variables (marriage and living with children) tend to have positive effects for men, but negative or null effects for women. In the equation for earnings, hours worked is always a prime predictor.

Results for Entrepreneurial Immigrants: Chinese and Koreans

The first two columns in every table provide results for the Chinese and Koreans, the most entrepreneurial of the immigrant groups studied in New York and Los Angeles. As shown in Table 5, self-employment does not enhance the odds of working last week - our indicator of whether they actually have jobs - for Chinese or Korean men or women in either metropolis.

There are some scattered effects of ethnic sectors, however. For example, working in the enclave has a positive effect for Korean men and women in New York (though not in Los Angeles).

Self-employment and being in ethnic sectors have more consistent and mostly positive effects on hours worked, particularly for men. The simplest case is that of Chinese men in New York. Self-employment itself increases hours worked by over 400 hours (about a 20% increase over the mean). In addition, working in any type of ethnic sector adds 200-500 hours for these men. The results are more selective for Chinese men in Los Angeles because the effects are interactive. For them, the combination of self-employment and being in the enclave or entrepreneurial niche strongly increases hours worked, but there is a negative effect (for both owners and workers) of being in the employment niche compared to being a worker in the mainstream economy. Other positive effects are found for Korean men from working in the enclave or entrepreneurial niche in New York and from being self- employed and in the enclave (and especially of combining both attributes) in Los Angeles. The models for Chinese and Korean women display a less clear pattern, but there is evidence of a positive effect of being in enclave sectors for Korean women in New York and of the combination of self-employment and working in the enclave for Korean and Chinese women in Los Angeles.

Hence we see these findings as a mirror of our results for Cubans: business ownership and the ethnic economy may not increase the odds of working, but they increase working hours for those who do have jobs. In terms of working hours, there is support for Hypotheses 1, 3 and 4.

In contrast to these benefits in working hours, the results for (logged) annual earnings are largely neutral (for Koreans) or negative (for Chinese).

For Korean men and women in Los Angeles, there are no significant effects

(25)

TABLE 5 LOGISTIC REGRESSION MODELS PREDICTING WORKING LASTWEEK VS. NOT WORKING LAST WEEK Chinese Korean African American Puerto Rican Dominican New York - Males Self-employed 0.544 (0.322) -0.010 (0.469) -0.402 (0.129) -0.387 (0.152) -0.362 (0.247) Enclave -0.315 (0.162) 0.881 (0.346) 0.349 (0.102) b 0.323 (0.193) Entrepreneurial niche 0.831 (0.386) 0.418 (0.340) 0.287 (0.108) b 0.332 (0.291) Employment niche 0.820 (0.765) b 0.123 (0.096) -0.292 (0.117) -0.149 (0.134) Self x Enclave 0.211 (0.581) -0.240 (0.660) a b 0.364 (0.528) Self x Ent. Niche -0.331 (0.886) 5.828 (9.419) 0.460 (0.302) b 0.310 (0.512) Self x Emp. Niche a b 0.114 (0.449) a a Public Employee b b 0.344 (0.068) 0.397 (0.101) b Constant 2.828 (0.399) 2.073 (1.061) 0.613 (0.258) 0.561 (0.179) 1.428 (0.407) Likelihood ratio 67.3 (df= 18) 35.0 (df=16) 498.2 (df=19) 218.9 (df= 15) 67.7 (df= 18) N of Cases 2,638 843 10,000 5,761 2,339 New York - Females Self-employed -0.527 (0.325) 2.322 (1.235) -0.147 (0.230) 0.183 (0.246) -0.333 (0.371) Enclave 0.088 (0.180) 0.876 (0.292) 0.129 (0.074) b 0.216 (0.388) Entrepreneurial niche -0.557 (0.264) 0.725 (0.322) -0.029 (0.172) b -0.694 (0.443) Employment niche 0.646 (0.385) b 0.089 (0.074) -0.155 (0.142) -0.465 (0.112) Self x Enclave 0.734 (0.683) -2.442 (1.290) -0.035 (0.372) b a Self x Ent. Niche a b a b a Self x Emp. Niche a b -1.100 (0.439) a a Public Employee b b 0.280 (0.065) 0.302 (0.096) b Constant 1.991 (0.386) 0.251 (1.106) 0.029 (0.258) 0.515 (0.221) 1.250 (0.380) Likelihood ratio 34.1 (df=18) 29.9 (df=16) 321.9 (df=19) 166.7 (df=15) 58.5 (df=18) N of Cases 2,309 641 10,000 4,850 2,074

(26)

TABLE 5 (CONTINUED)

LOGISTIC REGRESSION MODELS PREDICTING WORKING LAST WEEK VS. NOT WORKING LAST WEEK

Chinese Korean African American Mexican Salvadoran

Los Angeles - Males

Self-employed 0.009 (0.278) 0.067 (0.438) 0.172 (0.133) 0.293 (0.217) -0.057 (0.269) Enclave -0.058 (0.199) -0.066 (0.326) 0.106 (0.101) -0.425 (0.176) -0.512 (0.236) Entrepreneurial niche 0.142 (0.163) -0.287 (0.256) 0.519 (0.136) b -0.148 (0.780) Employment niche 0.069 (0.391) b 0.028 (0.081) -0.119 (0.068) -0.052 (0.131) Self x Enclave -0.254 (0.462) 0.590 (0.605) 0.124 (0.477) 0.577 (0.412) a

Self x Ent. Niche -0.115 (0.365) 0.225 (0.511) a b a

Self x Emp. Niche a b -0.292 (0.319) -0.363 (0.275) -0.222 (0.533)

Public Employee b b 0.518 (0.085) b b

Constant 1.791 (0.388) 2.674 (0.950) 0.205 (0.458) 1.608 (0.155) 1.310 (0.606) Likelihood ratio 104.7 (df=18) 38.9 (df=16) 401.8 (df=19) 200.3 (df=16) 54.1 (df=18)

N of Cases 2,937 1,651 7,503 10,000 2,644

Los Angeles - Females

Self-employed -0.242 (0.327) -0.085 (0.460) 0.154 (0.195) 0.117 (0.142) 0.214 (0.279) Enclave -0.019 (0.169) 0.081 (0.237) 0.073 (0.082) -0.514 (0.283) -0.012 (0.159) Entrepreneurial niche -0.114 (0.155) -0.225 (0.196) 0.298 (0.152) b -0.032 (0.475) Employment niche -0.325 (0.380) b -0.143 (0.076) -0.226 (0.058) 0.331 (0.120) Self x Enclave 1.578 (0.797) 0.735 (0.561) a a (0.898) a Self x Ent. Niche 0.380 (0.434) 0.945 (0.558) a b

Self x Emp. Niche a b -0.301 (0.328) -0.004 (0.296) a Public Employee a b 0.330 (0.073) b b

Constant 1.888 (0.392) 2.436 (0.807) 0.048 (0.426) 1.361 (0.142) 1.716 (0.622) Likelihood ratio 97.1 (df=18) 29.9 (df=16) 279.6 (df=19) 267.5 (df= 16) 40.5 (df=18)

N of Cases 2,550 1,371 8,358 10,000 2,450

Notes: a The coefficients for these cells are omitted because they are based on less than 25 cases.

b This group not overrepresented in this sector. c Likelihood ratio for logistic regression model and R-square for OLS models.

(27)

TABLE 6 MULTIPLE REGRESSION MODELS PREDICTING ANNUAL HOURS WORKED (IN 100s) Chinese Korean African American Puerto Rican Dominican New York - Males Self-employed 4.177 (0.650) 1.894 (1.495) -0.433 (0.350) 1.242 (0.407) 1.084 (0.854) Enclave 2.283 (0.452) 2.172 (0.974) -0.053 (0.206) b 1.496 (0.563) Entrepreneurial niche 2.441 (0.745) 2.571 (1.047) 0.714 (0.231) b 0.129 (0.806) Employment niche 5.216 (1.558) b -0.278 (0.227) -0.032 (0.320) -0.101 (0.437) Selfx Enclave 0.183 (1.198) 2.211 (1.853) b 4.763 (1.566) Self x Ent. Niche 0.124 (1.503) 1.095 (2.250) 1.137 (0.669) b 1.630 (1.502) Self x Emp. Niche b 1.546 (1.002) Public Employee b b 0.391 (0.147) 0.018 (0.212) b Constant 20.086 (0.966) 21.597 (3.183) 17.292 (0.591) 18.208 (0.459) 16.509 (1.206) R2 0.053 0.059 0.013 0.014 0.035 N of Cases 2,282 716 9,562 4,619 1,811 New York - Females Self-employed 1.520 (1.011) 3.106 (2.181) -1.912 (0.561) -1.909 (0.655) 0.883 (1.415) Enclave -0.316 (0.504) 2.306 (1.153) 0.480 (0.151) b -2.553 (1.262) Entrepreneurial niche 0.716 (0.903) -0.494 (1.200) 0.934 (0.395) b -2.066 (1.718) Employment niche 1.479 (0.871) b -0.028 (0.169) 0.289 (0.432) -0.776 (0.420) Self x Enclave 3.298 (1.722) -1.204 (2.526) 0.550 (0.921) b a Self x Ent. Niche a a b a Self x Emp. Niche a b 3.812 (1.419) a Public Employee b b -0.270 (0.134) -0.634 (0.225) b Constant 21.607 (1.022) 16.596 (4.098) 18.878 (0.656) 17.275 (0.614) 17.060 (1.256) R2 0.013 0.037 0.010 0.015 0.014 N of Cases 1,974 476 10,000 3,781 1,455

(28)

TABLE 6 (CONTINUED)

MULTIPLE REGRESSION MODELS PREDICTING ANNUAL HOURS WORKED (IN 100S) Chinese Korean African American Mexican Salvadoran

Los Angeles - Males

Self-employed 1.062 (0.602) 2.291 (0.917) -0.236 (0.358) 0.916 (0.358) 1.912 (0.652) Enclave 0.037 (0.461) 1.918 (0.778) -0.523 (0.264) -0.968 (0.428) 0.575 (0.641) Entrepreneurial niche -0.255 (0.335) 0.206 (0.591) 0.520 (0.289) b -0.083 (1.796) Employment niche -3.105 (0.879) b -0.207 (0.228) -0.214 (0.133) 0.426 (0.301) Self x Enclave 3.677 (1.075) 2.974 (1.256) 0.959 (1.158) 0.477 (0.753) a

Self x Ent. Niche 1.951 (0.771) 1.096 (1.099) a b a

Self x Emp. Niche a b 1.547 (0.901) -0.729 (0.500) -1.120 (1.347)

Public Employee b b 0.273 (0.198) b b

Constant 19.744 (0.849) 17.445 (2.066) 17.667 (1.352) 18.077 (0.310) 17.201 (1.711)

R2 0.083 0.115 0.032 0.043 0.033 N of Cases 2,599 1,448 5,976 10,000 2,188

Los Angeles - Females

Self-employed -0.192 (0.842) 2.773 (1.538) -0.734 (0.473) 0.063 (0.340) -1.214 (0.870) Enclave -0.004 (0.405) -0.246 (0.774) -1.780 (0.187) -0.167 (0.872) 1.098 (0.540) Entrepreneurial niche 0.007 (0.356) -0.359 (0.652) 1.012 (0.325) b -1.684 (1.502) Employment niche -2.948 (0.835) b 0.104 (0.193) 0.965 (0.152) 0.121 (0.379) Self x Enclave 5.957 (1.364) 5.078 (1.784) a a (2.808) a Self x Ent. Niche 1.853 (1.075) 0.005 (1.762) a b a

Self x Emp. Niche a b 2.941 (0.905) 0.335 (0.766) a Public Employee b b -0.098 (0.162) b b

Constant 17.626 (0.913) 14.404 (2.069) 15.879 (1.374) 18.324 (0.341) 14.728 (1.770)

R2 0.046 0.108 0.031 0.020 0.017 N of Cases 2,117 1,056 6,752 10,000 1,835

Notes: a The coefficients for these cells are omitted because they are based on less than 25 cases.

b This group not overrepresented in this sector. 'Likelihood ratio for logistic regression model and R-square for OLS model.

(29)

TABLE 7 MULTIPLE REGRESSION MODELS PREDICTING LOGGED ANNUAL EARNINGS Chinese Korean African American Puerto Rican Dominican New York - Males Self-employed -0.190 (0.057) -0.086 (0.120) -0.273 (0.038) -0.088 (0.043) -0.022 (0.088) Enclave -0.447 (0.040) -0.252 (0.079) -0.063 (0.022) b -0.142 (0.058) Entrepreneurial niche -0.244 (0.065) -0.218 (0.084) 0.114 (0.025) b -0.027 (0.083) Employment niche -0.367 (0.136) b 0.093 (0.025) -0.110 (0.034) -0.148 (0.045) Self x Enclave 0.124 (0.105) 0.211 (0.149) b -0.215 (0.161) Self x Ent. Niche -0.088 (0.131) 0.038 (0.181) -0.189 (0.073) b -0.519 (0.154) Self x Emp. Niche b 0.415 (0.109) Public Employee b b 0.197 (0.016) 0.151 (0.022) b Constant 8.498 (0.092) 8.642 (0.264) 8.090 (0.067) 8.202 (0.056) 8.538 (0.130) R2 0.436 0.246 0.302 0.308 0.235 N of Cases 2,282 716 9,562 4,619 1,811 New York - Females Self-employed -0.231 (0.083) -0.240 (0.174) -0.225 (0.054) -0.389 (0.064) 0.049 (0.128) Enclave -0.371 (0.041) -0.263 (0.092) -0.051 (0.015) b -0.311 (0.114) Entrepreneurial niche -0.226 (0.074) -0.014 (0.095) 0.048 (0.038) b -0.328 (0.155) Employment niche -0.439 (0.071) b -0.137 (0.016) -0.120 (0.042) -0.066 (0.038) Self x Enclave 0.305 (0.141) 0.321 (0.201) -0.314 (0.089) b a Self x Ent. Niche a a a b a Self x Emp. Niche a b -0.282 (0.138) a Public Employee b b 0.041 (0.013) 0.058 (0.022) b Constant 8.466 (0.093) 8.782 (0.331) 8.114 (0.066) 7.963 (0.066) 8.231 (0.120) R2 0.532 0.292 0.345 0.412 0.347 N of Cases 1,974 476 10,000 3,781 1,455